How can a GUI be incorporated into the demo program?

[MikeTheWatchGuy]

Would it be helpful to have a GUI?

I am interested in adding one.

What I'm thinking would be helpful would be to be able:

• View the results in "realtime" as the images are being processed
• Enable changing the parameters in realtime so the effect can be immediately studied
• Show multiple output windows on one screen. This allows comparison, side-by-side.
• Control "playback". Pause, Resume, and Seeking (particularly useful if you're experimenting with parameters)

I have a YOLO demo running that enables you to do something similar. You can change the Threshold and Confidence values on the fly.

If someone can give me a rough idea for how a GUI can be incorporated, I'll give it a shot.

[YilmazOrkun]

Also I would like to deploy the model to an application, where it can be tested with live images from a camera.

I found a cpp-wrapper for torch, which I want to try. https://github.com/tuotuoxp/cpp-torch If that does not work my second approach would be to convert the model from torch -> pytorch -> ONNX and use caffe2 for deployment.

What do you think?